Bridging the intention-behavior gap in academic entrepreneurship: the impact of training on faculty and graduate student researchers

Despite the importance of technology innovation and commercialization to the economy, graduate students and faculty on the front lines of scientific discovery rarely experience the process of bringing new discoveries to society through entrepreneurship (Heitor et al., 2014; Huyghe and Knockaert, 2015; Åstebro et al., 2012). One reason for this is that most doctoral training remains rooted in scientific rigor with limited emphasis on business or professional skills, even though more than half of U.S.-trained scientists and engineers will work in the private sector (Opsomer et al., 2021). PhDs who choose to remain in academia have been described as having a “strong taste for science” (Roach and Sauermann, 2010, p. 423), characterized by a preference for basic research and the freedom to choose research projects, in contrast to the more applied research that occurs in startups and industry. In response, universities and federal agencies have developed initiatives to help foster a more entrepreneurial academic culture that places value on translating research into economic and social impact.

To bridge the gap between academic research and more practical applications that drive innovation and benefit society, universities create policies, programs, and infrastructure designed to encourage faculty and graduate students to explore the commercialization of their research (Etzkowitz, 2004; Gibb and Hannon, 2006; Wright and Phan, 2018). These initiatives include investing in technology transfer offices (TTOs); promoting entrepreneurship centers, incubators, and accelerators; establishing mentorship and networking opportunities; promoting awards and recognition; and offering incentives such as technology development grants, favorable licensing terms, and leaves of absence (Barr et al., 2009; Valdivia, 2013). More recently, there has even been growing support for recognizing commercialization and entrepreneurship activities as part of the faculty tenure and promotion process to legitimize them as valued academic contributions (Carter et al., 2021).

Over the years, delivering entrepreneurship training to academic researchers has also become a key element of these efforts (Boh et al., 2016; Souitaris et al., 2007). Typically, the type of training directed at faculty and graduate student researchers is very experiential, focused on evaluating innovations emerging from university labs, with a curricular focus on conducting market research, developing business models, crafting pitches, and connecting with entrepreneurship stakeholders (Duval-Couetil et al., 2021b; Siegel et al., 2003). Research has found that entrepreneurship education can inspire and motivate researchers, which is important to the success of early-stage startups (Baum and Locke, 2004). It can also positively influence colleagues and students in their immediate environments (Liñán and Fayolle, 2015). Involving academic scientists in entrepreneurial training also offers secondary benefits, including interdisciplinary collaboration, networking, and boundary spanning across different cultural and organizational settings (Klofsten et al., 2021).

Few studies have explored how training directly influences faculty and graduate student involvement in entrepreneurship (Klofsten et al., 2021; Nabi and Holden, 2008). Scholars emphasize the need for more research in this area, given the potential positive impacts of education on self-efficacy and entrepreneurial behaviors (Kubberød and Pettersen, 2017; Rauch and Hulsink, 2015). Studying this population, in particular, is interesting, given studies suggesting that entrepreneurship education designed for business students may not be as effective for those in engineering disciplines (Maresch et al., 2016), as well as evidence suggesting that traditional institutional approaches to technology commercialization may overlook “nuanced differences in the goals, motives, and experiences of academic scientists” (Hmieleski and Powell, 2018, p. 44). Furthermore, while considerable evidence shows that entrepreneurship education offers participants practical skills, role models, and expanded networks, there is mixed evidence that it increases the likelihood of starting new companies (Hayter et al., 2017; Åstebro and Hoos, 2021).

EI is a research construct commonly used to measure the impact of entrepreneurship education (Liñán and Fayolle, 2015; Nabi et al., 2017). EI is conceptualized as “the conscious state of mind that precedes action and directs attention toward entrepreneurial behaviors” (Moriano et al., 2012, p. 165). Many studies exploring the formation of EI draw on Ajzen’s theory of planned behavior (TPB; Ajzen, 1985). The TPB has been widely examined and posits that three elements provide the basis for the intention to perform a behavior: (1) attitude toward the outcomes of the behavior, (2) social/subjective norms, and (3) perceived behavioral control. Attitude toward the outcomes of the behavior is the degree to which an individual has a positive or negative judgment of the behavior at hand. Social/subjective norms refer to the degree to which individuals perceive their behavior as aligned with that of others. Perceived behavioral control defines the extent to which a target behavior is within one’s ability. The latter overlaps with Bandura’s concept of perceived self-efficacy, defined as one’s personal beliefs about their ability to succeed in performing a target behavior (Bandura, 1994; Liñán et al., 2011).

Studies, including meta-analyses and reviews, have found positive relationships between entrepreneurship education and EI (Bae et al., 2014; Lorz et al., 2013; Martin et al., 2013); however, other studies have found no relationship or even negative ones (Nowiński and Haddoud, 2019; Oosterbeek et al., 2010; Siu and Lo, 2013; Von Graevenitz et al., 2010). Scholars have pointed out factors that are likely the cause of these mixed outcomes, including differences in educational programs, their target populations, and the business contexts (Nabi et al., 2017, 2018; Newman et al., 2019; Schmutzler et al., 2019). Methodological differences, such as sample sizes, measurement instruments, and study design, are also factors contributing to the mixed results (Rideout and Gray, 2013; Yi and Duval-Couetil, 2022).

While the TPB has been widely used in entrepreneurship research to understand and predict EI, it has been criticized for its inadequate predictive capacity and limitations in addressing the many factors involved in entrepreneurial decision-making (Nabi et al., 2017). Many have pointed out that motivation, intuition, and context often drive outcomes more than intention alone (Krueger, 2000; Nabi et al., 2017). Although the TPB does not capture all of the complexities involved, it nevertheless offers a structured framework for examining how education can shape EI and possibly lead to entrepreneurial behavior.

Research applying the TPB to examine entrepreneurship education’s impact has primarily focused on the formation of EI, with relatively few studies exploring the relationship between intention and behavior (Kautonen et al., 2015). The intention-behavior gap describes the discrepancy between an individual’s intention to engage in a specific action, such as starting a company, and their failure to convert that intention into actual behavior (Van Gelderen et al., 2015). Kautonen et al. (2015) highlighted the shortage of studies addressing this gap, noting that “entrepreneurship is about actions rather than mere intentions, and the extent to which entrepreneurial intentions translate into action defines the relevance of intention research” (p. 656). Others point out that EI has limited value as a construct if it cannot be connected to future entrepreneurship behavior through longitudinal research (Yi and Duval-Couetil, 2022).

Within the context of academic entrepreneurship, evidence points to a considerable intention-action gap (Hossinger et al., 2020). In a study of academic scientists, Fritsch and Krabel (2012) found that while 28% found it attractive or highly attractive to start a company, only 3.2% actually engaged in startup activity. Many academics are reluctant to engage in entrepreneurship for many personal and professional reasons. A common barrier is the lack of business knowledge, leaving many researchers unprepared to navigate commercialization (Clarysse et al., 2011). Some faculty view commercialization as distracting from their core academic teaching and research responsibilities (Perkmann et al., 2013). Others believe integrating business with research could compromise scientific integrity (Lam, 2010). Additionally, the tenure and promotion system tends to reward publishing more than patenting (Baldini, 2006). Institutional factors also play a role, as university and TTO policies are often perceived as bureaucratic and complex (Siegel et al., 2007). Finally, some are unwilling to take financial and professional risks (Haeussler and Colyvas, 2011).

Entrepreneurship education is considered to play an important role in bridging the intention-behavior gap by equipping participants with the knowledge, skills, and confidence needed to translate intentions into actual entrepreneurial behaviors (Tran et al., 2024). While aspiring entrepreneurs may strongly intend to start a business, barriers such as a lack of resources, self-efficacy, or risk perceptions can prevent them from taking action (Van Gelderen et al., 2015). Entrepreneurship education mitigates these barriers through experiential learning, mentorship, and interaction with entrepreneurs (Souitaris et al., 2007). Through opportunities to test ideas, receive feedback, and develop entrepreneurial problem-solving skills, education better prepares participants for the realities of venture creation.

Given the number of factors that can influence one’s decision to pursue entrepreneurship, considering the influence of moderating variables can offer deeper insights into how training bridges the intention-behavior gap. Moderators affect the strength or direction of a relationship between independent and dependent variables, in this case, between EI and entrepreneurial behavior (Baron and Kenny, 1986). Within the context of entrepreneurship, they can include demographics (e.g. age, gender, cultural background), individual traits (e.g. personality, prior experience), and contextual influences (e.g. institutional environments, support networks). Curricular and program attributes, such as instructional approach and training duration, can also play a role. Other influences include motivation, attitudes toward entrepreneurship, psychological influences, and external factors like industry opportunities and economic trends (Kautonen et al., 2015; Nabi et al., 2017; Yi and Duval-Couetil, 2018).

For academic entrepreneurs, research has pointed to several moderating variables influencing the intention-behavior gap. Gender and family background are influential, given that male academics and those with entrepreneurial parents are more likely to participate in patenting, industry collaboration, and venture creation (Miranda et al., 2017a, b). Age exhibits positive and negative effects associated with commercialization activities (Neves and Brito, 2020). In terms of institutional factors, the influence of university administrators and department chairs can be more important than that of colleagues (Brettel et al., 2013; Hossinger et al., 2020). TTO support has been found to have both positive and null effects (Bourelos et al., 2012; Brettel et al., 2013; Clarysse et al., 2011; Fini et al., 2009; Wu et al., 2015). External factors, such as low salaries, can also play a role; for example, 90% of academic scientists in Hungary created ventures to supplement their incomes (Novotny, 2017).

Overall, a lack of research supports the assumption that EI leads to actual behavior (Bae et al., 2014). Studies designed to link intention to behavior are difficult to conduct as they involve tracking participant attributes and their startup activities over time, which is resource-intensive in terms of data collection and prone to low survey response rates, biased responses, and the influence of confounding variables (Bauer, 2004; Yi and Duval-Couetil, 2022). Overcoming these issues requires a large-scale initiative with consistent program elements and significant resources dedicated to participant tracking and survey administration.

This study’s overarching research question is: How does entrepreneurship training influence the entrepreneurial intention and behavior of academic researchers, particularly in relation to startup creation and the pursuit of funding? Our conceptual model is based on the TPB, which provides a structured framework for examining how entrepreneurial training impacts academic researchers' EI and behaviors, as well as the role of moderating variables in this process. First, we test the direct effects of training on EI and whether post-training EI predicts entrepreneurial behavior. Second, we test the moderating effects of three factors pertinent to academic entrepreneurship: team role, project potential, and prior entrepreneurial experience. The specific hypotheses and their groupings are described in detail below.

Standard EI outcome measures for entrepreneurship education and training targeting academic researchers include assessments of commercialization readiness, participants’ confidence in transitioning their research to market applications, whether they intend to pursue commercialization through startup activity, and whether they intend to seek funding to continue to develop and commercialize their technologies (Markman et al., 2008). Typical measures of entrepreneurship behavior include outcomes such as founding a company, pursuing follow-on funding and investment, and career shifts involving taking on entrepreneurial roles in startups.

These measures capture distinct yet interrelated aspects of the entrepreneurial process. Making determinations about the commercialization readiness of particular technologies is particularly relevant in academic entrepreneurship as this activity is a desired outcome of training targeting faculty and graduate student researchers (Barr et al., 2009). Startup intention is a core construct in entrepreneurship research as it reflects an individual’s motivation and commitment to launch a new venture (Nabi et al., 2017). Securing financial resources is another critical step in venture creation, as one’s willingness to seek funding indicates entrepreneurial behavior (Liñán and Fayolle, 2015), and evidence suggests that entrepreneurship training can increase one’s confidence in pursuing external funding (Munir et al., 2022). Finally, the ultimate goal of entrepreneurship programming is to create startup companies that are able to bring new innovations to the market. Therefore, we propose the following hypotheses representing the direct effects of training on EI and behavior:

An individual’s role on an entrepreneurial team is an important factor to consider when examining shifts in intention and behavior because entrepreneurial ventures are often founded and led by teams rather than individuals (Beckman, 2006). In academic entrepreneurship, faculty and graduate students are recognized as distinct subgroups assuming particular roles on entrepreneurial teams (Hayter et al., 2017). While most prior research focuses on faculty, a small but growing body of literature examines the critical role of graduate students and postdoctoral researchers in startup activities (Boh et al., 2016; Hayter et al., 2017). Given their in-depth knowledge of emerging technologies, they are considered to play an essential role in the early stages of commercialization and startup activity (Bramwell and Wolfe, 2008; Hayter et al., 2017). Research has found them to be more likely than faculty to create academic startups (Åstebro et al., 2012), and to be involved in successful ones (Rasmussen et al., 2014). In a study of eight universities, Boh et al. (2016) found that of 47 academic startups, Ph.D. students and postdocs were involved in 36 (77%), and of those, 11 (23%) were established without faculty involvement.

Research suggests that graduate students have greater EI than faculty members (Hayter et al., 2017; Huyghe and Knockaert, 2015). Their age and career stage make them more open to diverse employment opportunities and enthusiastic about pursuing entrepreneurship. They also face lower opportunity costs when pursuing entrepreneurship because they do not yet have full-time positions (Cohen et al., 2020). However, faculty are also critical to startup creation, given their greater scientific expertise, which they can use to refine technologies and attract investors (Fuller and Rothaermel, 2012). These distinctions highlight potential group differences in EI and entrepreneurial behaviors and the importance of considering team roles when assessing the impact of training. Therefore, we propose the following hypotheses:

Another factor impacting EI and behavior is whether a technology project is ready for commercialization. Entrepreneurship training for academic researchers is typically focused on using emerging technology as the “platform for learning” (Barr et al., 2009, p. 372), where a desired outcome is establishing whether it is “ready for commercialization” or “not yet ready for commercialization” (NSF, 2021; Laukkanen, 2000). This contrasts with training that is more focused on general entrepreneurship principles for non-scientific audiences (Barr et al., 2009). This curricular approach creates a situation where EI and entrepreneurial behavior may be influenced more by the market potential of a specific technology project (project-dependent factors) than by individual-level factors such as motivation or self-efficacy (person-dependent factors). If this is true, then effective instruction alone may not be sufficient to enhance EI. Instead, outcomes depend more on the market readiness of the technology being evaluated. Therefore, we propose the following hypotheses:

Prior experience is an individual-level factor related to entrepreneurial self-efficacy (ESE), which is considered a predictor of EI (Miao et al., 2017; Newman et al., 2019). Research supports a relationship between prior industry or entrepreneurial experience and involvement in academic entrepreneurship (Criaco et al., 2014; Rasmussen et al., 2014; Shane et al., 2015). Experiences with patenting (D'Este et al., 2012; Kolympiris and Klein, 2017), product development (Karlsson and Wigren, 2012), and startups (Abreu and Grinevich, 2013; Muscio et al., 2016; Scholten et al., 2015), have been shown to predict entrepreneurial behavior. For example, a study of academics in Norway found that work experience outside of academia positively affected the creation of startups, patenting, licensing, and industry collaboration (Gulbrandsen and Thune, 2017). Other studies support the role of prior experience in startup creation (Bergmann, 2017; Miranda et al., 2017a, b; Obschonka et al., 2015). These findings highlight that prior experiences, not typically required for academic work, provide academic scientists with specialized knowledge that can contribute to university commercialization activities, and this specialized human capital can enhance the performance of startups (Toole and Czarnitzki, 2009). This suggests that graduate students and faculty with relevant experience will likely be better equipped to translate EI into actual entrepreneurial behavior. Therefore, we propose the following hypotheses:

Our sample consists of faculty and graduate students who participated in the National Science Foundation (NSF) Innovation Corps (I-Corps) program, a nationwide training program launched in 2012 to stimulate entrepreneurial activity at US universities. The I-Corps training curriculum is based on the Lean Launch methodology (Blank and Dorf, 2012), which focuses on an activity known as “customer discovery” that involves conducting 100 interviews with potential customers to assess the market potential of new technologies and the development of viable business models (VentureWell, 2023). The I-Corps program has several stated objectives, including helping academic researchers bring their innovations to market, equipping them with an entrepreneurial mindset, and building entrepreneurial ecosystems that can contribute to job creation and economic growth (National Science Foundation, 2012).

Through a competitive process, teams of three individuals apply to the I-Corps program with a specific technology commercialization project in development. Members of the team include a Principal Investigator (PI), typically a faculty member able to provide deep technical expertise; an Entrepreneurial Lead (EL), usually a graduate student or postdoctoral researcher who is expected to lead the project; and an I-Corps Mentor (IM), a volunteer business advisor who provides expertise (Blank and Engel, 2016). When the data were collected, the national I-Corps program involved in-person meetings at the beginning and end of a seven-week training period with five weeks of online instruction. Selected teams received $50,000 in grants from the NSF to support travel expenses associated with conducting customer discovery interviews and some modest salary support for the EL (National Science Foundation, 2021).

Data were collected through surveys distributed by the nonprofit organization VentureWell on behalf of the National Science Foundation (NSF). These data are publicly available as “Public-Use Data, Release 4.0” (VentureWell, 2023). Participants completed three online surveys distributed via email: one before the program (“pre-course survey”), one immediately following the program (“post-course survey”), and one 9–16 months after completion (“longitudinal follow-up survey”). The longitudinal survey was administered in September of each year to all participants in the previous year, with an average period of 13 months after program completion. VentureWell developed the surveys internally for program evaluation, and information on convergent and discriminant validity was not available. See Table 1 for the survey questions included in this study.

Surveys were sent to 78 cohorts of I-Corps teams that participated in the national program between 2012 and 2019. Although the dataset includes responses from “I-Corps Mentors” (IM) responses, they were excluded from our analyses to maintain a specific focus on academic researchers. This resulted in a sample of 3,455 participants (EL = 1,846, PI = 1,609). Examining the integrity of the data revealed that there were 1,191 participants with complete pre-survey data (EL = 630, PI = 561), 1,198 participants with complete post-survey data (EL = 638, PI = 560), and 411 participants completed the longitudinal follow-up survey (EL = 203, PI = 208). Demographic data were not available. See Table 2 for the descriptive statistics.

The following three variables represented changes in EI over time. Survey items representing these variables were included in both the pre- and post-course surveys, and 5-point response scales were used (Table 1).

Startup intention was represented by a single survey question, “I am interested in starting my own company,” with response choices ranging from “Totally disagree” (1) to “Totally agree” (5).

Funding intention was represented by two survey questions: “I will apply for SBIR funding for my technology in the next 12 months” and “I will seek other funding or investment for my technology within the next 12 months,” with response choices ranging from “totally disagree” (1) to “totally agree” (5). SBIR is the acronym for the Small Business Innovation Research, a federal program that invests in technology development for startups. The maximum value among the responses to these two questions formed a single variable, thereby capturing the strongest expressed intention toward securing funding.

Commercialization readiness was assessed using four survey items, averaged to create a composite scale (α = 0.750 for the pre-survey; α = 0.835 for the post-survey). These items assessed participants' progress in developing a viable commercialization strategy. Specifically, they indicated whether they had identified a commercialization path, developed a scalable business model, defined a minimum viable prototype, and identified and validated a market. There was no previously validated scale for this measure, which is fairly common in entrepreneurship research where new scales are often developed for specific studies without prior validation (Åstebro and Hoos, 2021). Therefore, we conducted a confirmatory factor analysis (CFA) to assess the validity of this construct. The standardized factor loadings and communalities are presented in Table 3. All loadings were statistically significant (p < 0.001), and most exceeded 0.70, indicating strong associations between the observed variables and their corresponding latent constructs. The survey item “I have defined a minimum viable prototype for my product (CR3)”, however, showed slightly weaker contributions, with standardized loadings of 0.524 (pre) and 0.662 (post). Communalities ranged from 0.274 to 0.692, suggesting that the latent constructs explained between 27.4 and 69.2% of the variance in the observed variables. Our analysis demonstrates that the scale performs reliably and is suitable for examining commercialization readiness in a large-scale setting.

The following two categorical variables represented participants’ self-reported entrepreneurial behavior, with data collected via the longitudinal follow-up survey (Table 1).

Startup behavior was assessed using one question on the “timing of business established in relation to completion of the course,” with three response options: business established before, during, and after the course. We coded “before” and “during” the course as 0, and “after” as 1.

Funding behavior was assessed using a single survey item, “After the I-Corps course ended, did you seek any additional funding (e.g. a Small Business Innovation Research/SBIR award, equity financing from a venture capitalist, state grant, or personal investment)?” Responses were coded as a 0/1 dichotomous variable.

Moderators consisted of three variables that could potentially identify participant subgroups exhibiting distinct changes in EI and behavior.

Team role was reported as either EL (Entrepreneurial Lead), usually graduate students or postdoctoral researchers, or PI (Primary Investigator), usually faculty.

Project potential was represented by a single survey question: “Which of the following best describes the current status of your project,” which was associated with four response options: “Our research has commercial potential (Go decision),” “Our research is not yet ready for commercialization (No-go decision),” and “We are not yet sure whether our research is commercializable (Uncertain),” and “None of the above (please describe status in your own words below).” To preserve adequate cell sizes for analysis, this variable was converted into a dichotomous dummy coded measure, where “1” represented a “go decision” and “0” represented a “no-go decision” combined with the responses “uncertain” and “other.”

Prior experience was derived from the sum of a checklist of 14 survey questions, each assessing different types of entrepreneurial activities in which participants had engaged (Table 2). These questions collected open-ended numeric responses, asking how many times respondents participated in each activity. Due to the extreme skew of these numeric estimates, we instead calculated the total number of activity types in which each participant engaged. Therefore, this variable ranges from 0 to 14 (M = 5.87, SD = 3.47).

To test our hypotheses, we conducted a series of statistical analyses. First, we assessed the direct effects of entrepreneurship training on EI using paired-sample t-tests to compare pre- and post-training scores on startup intention (H1a), funding intention (H1b), and commercialization readiness (H1c). Next, we employed two logistic regression analyses to examine whether post-training EI predicted startup behavior (H2a) and funding behavior (H2b). We then conducted repeated measures ANOVAs to analyze the moderating effects of team role (H3a-H3c), project potential (H5a-H5c), and prior experience (H7a-H7c). Additionally, the moderating effects of team role on startup behavior (H4a) and funding behavior (H4b) were included as interaction terms in the logistic regression models. The same method was used to examine the moderating effects of project potential (H6a-H6b) and prior experience (H8a-H8b).

Most of the survey data used in this study were in Likert Scale format. Although there is ongoing debate regarding the treatment of Likert scale variables as numeric data, it is a widely accepted practice when the underlying construct is continuous and the intervals between response categories are approximately equal. This approach allows for the application of powerful parametric statistical techniques, such as ANOVA and regression analysis (Carifio and Perla, 2007; Norman, 2010), which can detect subtle differences between groups or over time that might be overlooked if the data were treated as purely ordinal (Sullivan and Artino Jr, 2013). In this study, using repeated measures ANOVA was particularly advantageous as it accounted for within-subject variability over time and incorporated moderating variables, enabling us to examine whether changes in EI were consistent across different participant groups and influenced by the moderators.

Paired sample t-tests detected a small but statistically significant decrease in in startup intention (M_pre = 4.33, SD_pre = 0.877, M_post = 4.28, SD_post = 0.932, t = −2.164, p = 0.015). However, the effect size was negligible (d = −0.031), suggesting limited practical significance. Therefore, H1a was not supported, meaning that training did not significantly affect the intention to pursue a startup. In contrast, funding intention increased significantly from pre-to post-training (M_pre = 4.35, SD_pre = 0.750, M_post = 4.42, SD_post = 0.839, t = 2.962, p = 0.003), providing support for H1b. Similarly, commercialization readiness showed a large and statistically significant increase (M_pre = 3.14, SD_pre = 0.714, M_post = 3.79, SD_post = 0.799, t = 30.717, p < 0.001), supporting H1c and meaning that training increased participants’ perceptions of commercialization readiness. See Tables 4 and 5 for full descriptive and inferential statistics.

Logistic regression was conducted to examine whether post-training EI predicted the likelihood of engaging in startup behavior. While the coefficients for all three predictors were positive: startup intention (B = 0.22, p = 0.087), funding intention (B = 0.243, p = 0.110), and commercialization readiness (B = 0.210, p = 0.143), none reached conventional levels of statistical significance. Therefore, H2a was not supported, suggesting that post-training entrepreneurial intentions alone were not strong predictors of startup behavior in this model (see Table 9). In terms of funding behavior, funding intention was the only statistically significant predictor (B = 0.471, p < 0.001), indicating that participants with higher funding intention were more likely to seek funding following the training. In contrast, startup intention (B = 0.168, p = 0.120) and commercialization readiness (B = −0.119, p = 0.404) were not significant predictors. These results partially support H2b, with only funding intention significantly predicting funding behavior (see Table 10).

To assess whether the effect of training on EI varied by team role (graduate student vs faculty), we conducted a series of repeated measures ANOVAs (Tables 6–8). For startup intention, the interaction between time (pre- and post-training) and team role was statistically significant (F(1, 1,622) = 7.648, p = 0.006, η_p² = 0.005), indicating that changes in startup intention differed by role. However, simple effects analyses showed that graduate students showed a slight decrease, whereas faculty showed a slight increase. Therefore, H3a was not supported. For funding intention, the interaction was not statistically significant (F(1, 1,776) = 1.162, p = 0.281, η_p² = 0.001), suggesting that the change in funding intention over time did not differ between faculty and graduate students. Therefore, H3b was not supported. However, a significant between-subjects effect was observed (F(1, 1,776) = 24.025, p < 0.001, η_p² = 0.013), indicating that faculty and graduate students still differed in their overall level of funding intention. In other words, team role matters, just not as a moderator of change. For commercialization readiness, the interaction was statistically significant (F(1, 1,800) = 9.499, p = 0.002, η_p² = 0.005), suggesting that team role moderated the effect of training on commercialization readiness. Therefore, H3c was supported.

For startup behavior, the interaction between startup intention and team role was not significant (B = 0.000, p = 0.995), indicating that the relationship between intention and startup behavior did not differ between faculty and graduate students. Therefore, H4a was not supported. For funding behavior, the interaction between funding intention and team role was also not significant (B = −0.016, p = 0.702), suggesting no moderating effect of team role. Therefore, H4b was not supported (see Tables 9 and 10).

For startup intention, a significant interaction was found between time and project potential (F(1, 1,458) = 45.916, p < 0.001, η_p² = 0.031), indicating that participants working on high-potential projects showed a greater increase in startup intention from pre-to post-training. Therefore, H5a was supported. Similarly, the interaction for funding intention was statistically significant (F(1, 1,602) = 59.265, p < 0.001, η_p² = 0.036), with greater increases observed among those with high-potential projects. Therefore, H5b was supported. For commercialization readiness, the interaction was particularly strong (F(1, 1,625) = 206.897, p < 0.001, η_p² = 0.113), again indicating that participants with high-potential projects demonstrated substantially higher levels of commercialization readiness. Therefore, H5c was supported. In all three models, significant between-subjects effects were also observed (p < 0.001), showing that participants with high-potential projects reported consistently higher levels of EI than those with lower-potential projects, regardless of time point (Tables 6–8).

Project potential significantly moderated the relationship between EI and both types of entrepreneurial behavior. For startup behavior, the interaction between startup intention and project potential was statistically significant (B = 0.216, p < 0.001), indicating that the positive association between startup intention and startup behavior was stronger for participants working on high-potential projects. Therefore, H6a was supported. For funding behavior, the interaction between funding intention and project potential was also significant (B = 0.190, p < 0.001), suggesting that participants with high-potential projects were more likely to translate funding intention into funding behavior. Therefore, H6b was supported (Tables 9 and 10).

For startup intention, the interaction between time and prior experience was not statistically significant (F(14, 1,293) = 0.519, p = 0.924, η_p² = 0.006), indicating that changes in startup intention did not differ based on levels of prior experience. Therefore, H7a was not supported. The interaction effect for funding intention was also not significant (F(14, 1,388) = 0.828, p = 0.639, η_p² = 0.008), suggesting that prior experience did not moderate the change in funding intention. Therefore, H7b was not supported. Similarly, there was no significant interaction for commercialization readiness (F(14, 1,408) = 0.367, p = 0.984, η_p² = 0.004), providing no support for H7c. However, in all three models, the between-subjects effects were statistically significant (p < 0.01), indicating that participants with greater prior experience consistently reported higher levels of EI across time points. In other words, prior experience had a significant direct effect, but not a moderating effect (see Tables 6–8.)

For startup behavior, a statistically significant interaction was found (B = 0.013, p = 0.012), indicating that the relationship between startup intention and startup behavior was stronger for participants with more prior experience. Therefore, H8a was supported. However, for funding behavior, the interaction was not statistically significant (B = 0.007, p = 0.173), providing no support for H8b (see Tables 9 and 10).

The purpose of this study was to examine the impact of training on the entrepreneurial intention and behavior of academic researchers, focusing on participants in the NSF I-Corps program, a national initiative designed to foster entrepreneurial activity at US universities and considered by many to be the “gold standard” in academic entrepreneurship training. The unique size and scale of the program, as well as its standardized curriculum, allowed us to minimize the effects of participant, curricular, and contextual differences, which are often cited as sources of variability in entrepreneurship education impact studies. We used the theory of planned behavior as a model to assess how training helps bridge the significant intention-behavior gap found among faculty and graduate student researchers. We also examined how individual and contextual variables interact with training to influence outcomes.

As stated, EI is a well-established precursor to entrepreneurial behavior, and many studies evaluate the impact of education and training solely by examining changes in EI given that entrepreneurship behavior can be difficult and costly to track. In our first set of hypotheses, we used three distinct measures of intention to assess pre- and post-training changes in EI—startup intention, funding intention, and commercialization readiness. We also included three moderating variables—team role (graduate students vs faculty), prior experience, and perceived project potential—to examine the factors influencing EI. When moderating variables were excluded from the analyses, our results supported the hypothesis that participation in I-Corps training increased commercialization readiness and funding intention, indicating that the program helped participants assess the commercialization potential of their technologies and motivated them to pursue funding. Interestingly, the hypothesis that I-Corps would increase startup intention was not supported, indicating that participants, on average, did not exhibit greater interest in starting their own companies after the training.

Including moderating variables in our analyses uncovered significant group differences that were previously masked. Specifically, startup intention decreased among graduate students but remained stable among faculty. This was a surprising finding, given prior research suggesting that graduate students are better suited to become academic entrepreneurs given their career and life stages (Rasmussen et al., 2014; Åstebro et al., 2012). This outcome may be related to the I-Corps curricular approach, which expects graduate students to conduct approximately 100 market research interviews during training, giving them a very realistic understanding of the challenges involved in practicing entrepreneurship. It is also possible that their entrepreneurial aspirations were closely tied to the market potential of the specific technologies they evaluated rather than being a more general career goal. Stability in faculty EI from pre-to post-training is less surprising given the well-documented mix of personal and professional barriers that prevent faculty from pursuing entrepreneurial activity (D’Este and Perkmann, 2011; Hossinger et al., 2020; Lam, 2011; Miller et al., 2018). Given their established careers, it is less likely that they are affected by project-specific outcomes. More research is necessary to understand how specific curricular approaches affect these post-training outcomes.

A very positive outcome of the I-Corps curricular approach is that perceptions of commercialization readiness increased, which is a well-established goal of entrepreneurship education directed at faculty and graduate student researchers. The program’s strong emphasis on market research and business model development likely plays an important role in this shift. As expected, participants with higher potential (“go” projects) showed more positive trends in EI. Not surprisingly, perceptions of project potential intensified differences in startup and funding intention, with “go” projects showing an increase in startup intention and “no-go” projects showing a decrease.

Interestingly, perceptions of commercialization readiness increased for both “go” and “no-go” projects. These findings suggest that I-Corps effectively helps participants develop realistic assessments of their projects' commercialization potential, even when the outcome is negative. This can be a valuable program outcome even for a “no-go” project if it enables faculty to pivot their research in a more market-facing direction. A benefit of learning is to evaluate technologies through customer discovery is that the knowledge and mindset acquired through this training can be applied to other technology transfer activities, industry engagement, and even teaching (Duval-Couetil et al., 2021a).

To examine the relationship between EI and the actual founding of a company, we examined whether post-training startup intention, funding intention, and commercialization readiness predicted startup behavior approximately one year after the training. We also tested the moderating effects of team role, prior experience, and perceived project potential. For startup behavior, the interaction effects between startup intention and both project potential and prior experience were significant predictors, while startup intention alone was not. This suggests that merely having a strong startup intention post-training was insufficient to predict actual startup behavior.

The positive link between funding intention and actually pursuing funding approximately one year after the training was stronger for participants working on projects with high project potential. Academic researchers’ strong interest in securing funding is not surprising given the professional norms in science and engineering, which require them to continually secure external grants to sustain research programs, support graduate students, and get tenure (Hossinger et al., 2020; Iorio et al., 2017). I-Corps is a unique training program in that participation is accompanied by NSF grant funding, which to some degree “legitimizes” it as a scholarly activity, potentially contributing to its popularity. Its positioning as a pathway for obtaining additional federal funding through initiatives such as SBIR grants also enhances its appeal. In the future, studies should distinguish between forms of startup funding (e.g. public versus private) when collecting data on startup activities from academic entrepreneurs to provide more insight into their scholarly versus entrepreneurial goals.

The strong association between prior experience and all measures, except for funding behavior, remained stable over time, indicating that individuals with greater experience were consistently more inclined toward startup and funding intentions and behaviors. Additionally, they exhibited a more refined ability to assess their projects' commercialization potential, likely benefiting from their accumulated knowledge and exposure to entrepreneurial processes. Our findings align with research underscoring the predictive value of prior experience in university commercialization activities. Industry engagement, patenting, product development, consulting, and other non-academic work have all been found to positively affect entrepreneurial behaviors (Criaco et al., 2014; Rasmussen et al., 2014; Shane et al., 2015).

Applying the TPB to the academic context revealed notable instances of the “intention-behavior gap,” with several findings contradicting our initial hypotheses. Notably, participation in entrepreneurship training did not lead to a significant increase in startup behavior, likely due to the complex mix of personal and professional factors influencing academics’ interest in commercialization discussed previously. However, this does not diminish the broader impact of I-Corps on startup creation and university culture. While startup formation is often used as a primary metric for assessing the impact of entrepreneurship education and training, it can overlook other valuable outcomes. For instance, startups emerging from I-Corps may launch more quickly or more successfully. Participants may also apply the training to commercialize technology through licensing agreements or industry partnerships rather than by founding ventures themselves. Future research should develop metrics that can account for technology transfer efficiency, depth of industry engagement, and the long-term impacts of entrepreneurship training on academic culture.

Given the strong predictive and moderating effects of a particular technology’s market potential, further research should explore its influence on entrepreneurial decision-making. Critical questions are: What factors shape the assessment of a project’s commercial potential? To what extent does market research inform these? How do the researchers and business mentors influence these evaluations? And, does training enable participants to assess project potential at an earlier stage or with greater precision, thereby improving decision accuracy and saving time? Understanding these mechanisms could lead to training refinements, ensuring they are better suited to participants and the technologies being evaluated. If certain training elements prove effective, future research should examine how best to integrate them into university education, research, and commercialization activities to enhance technology evaluation, entrepreneurial readiness, and innovation impact.

The impact of prior experience on entrepreneurial outcomes also presents a promising area for further research. While this study examined its general relationship with EI and behavior, more detailed analysis is needed to identify which specific types of prior experience are most influential for different entrepreneurial activities. For instance, to what extent does experience with market research have a stronger effect on commercialization readiness? Does prior startup leadership experience influence actual business formation? And, how do these factors interact differently for faculty and grad students? Clarifying these distinctions could inform the design of entrepreneurship training to align with participants’ backgrounds, learning needs, and institutional objectives.

Our findings suggest that training alone is insufficient to bridge the intention-behavior gap among academic researchers. Their stronger preference for pursuing funding rather than startups suggests that entrepreneurship initiatives and incentives must align with the priorities of academic scientists. Beyond education, universities can influence EI by offering funding support, facilitating industry connections, and implementing policies that encourage participation (Hossinger et al., 2020). Access to research grants, startup capital, venture funding, and entrepreneurial talent further supports technology commercialization involvement. Most importantly, both faculty and graduate students must balance traditional academic responsibilities with the demands of entrepreneurship activity. Structural accommodations, such as course releases, flexible timelines, or additional resources, could enable them to engage in entrepreneurial activities without compromising their academic commitments.

There is also an opportunity to align training objectives with participant and institutional characteristics. Our findings indicate that individuals with relevant prior experience demonstrate stronger intentions to pursue startups. Additional factors associated with higher participation include high scientific productivity (Clarysse et al., 2011), a focus on applied research (Hassinger), and a supportive departmental culture (Bercovitz and Feldman, 2008; Huyghe and Knockaert, 2015). Understanding these characteristics can help universities identify individuals most likely to engage in entrepreneurship and better allocate resources to achieve institutional strategic objectives (e.g. short-term startup activity versus longer-term culture change).

Given the significance of prior experience, universities should consider integrating more industry engagement into doctoral education if the goal is to graduate more entrepreneurial scientists and academics. Doing so can involve internships, collaborative research, or mentorship programs that expose students to real-world challenges and commercialization opportunities. Additionally, our results suggest that incorporating experiential learning elements into research training, such as “go/no-go” assessments, can help identify promising innovations. However, academic departments and faculty advisors must support these entrepreneurial goals, given the influence of norms and culture on perceptions of appropriate entrepreneurial behavior and acceptable career paths (Bercovitz and Feldman, 2008).

Finally, while our study focuses on the short-term outcomes up to approximately one year after training, it is crucial to consider the longer-term effects of entrepreneurship training on faculty, graduate students, and university innovation ecosystems. Policymakers and university administrators should track both immediate and sustained impacts to assess the broader influence of these programs on academic culture. Given the complexity and length of the entrepreneurial process, comprehensive evaluation metrics are needed. While many institutions attempt to track these outcomes, it is challenging given the resources and effort required to do it systematically.

This study contributes to the literature by examining the theory of planned behavior to examine the “intention-behavior gap” in academic entrepreneurship. Our findings confirm the existence of this gap for startup activity and highlight the complex interplay between individual-level differences, curricular approaches, and project characteristics. Given the resources being invested in university commercialization initiatives, it is important to continue identifying factors that influence participation and success, particularly for specific subgroups of participants. While short-term training outcomes are valuable, the broader objective of cultivating an entrepreneurial workforce and fostering innovation ecosystems requires long-term evaluation. Future research should focus on refining impact metrics that capture long-term entrepreneurial engagement, technology commercialization pathways, and shifts in academic culture. By deepening our understanding of these dynamics, universities and policymakers can design more effective interventions to bridge the intention-behavior gap and enhance the role of academic entrepreneurship in driving innovation.

Our findings should be interpreted within the context of several methodological limitations. First, I-Corps is a selective program, and EI is high for all participants entering the program, which may cause a ceiling effect. Second, the intention and behavior survey questions were not completely aligned; for example, the intention question relates to “my own company,” and the behavior question relates to a company based on the I-Corps project. Third, missing data could introduce potential bias in the analysis or affect the generalizability of the findings. Finally, the study measured prior experience representing a broad range of activities. Future analyses of the effects of prior experience would benefit from a more granular examination of the predictive power of specific types.

Finally, numerous contextual and individual factors influencing training outcomes warrant further investigation. At the program level, key influences include instructors, team dynamics, and the diversity of participant cohorts. At the university level, factors such as culture, commercialization infrastructure, and regional location can significantly impact entrepreneurial engagement. Beyond the university, broader influences, including the nature of emerging technologies, industry-specific opportunities, and prevailing economic conditions, also play a critical role. Most importantly, personal circumstances, such as financial resources, family obligations, and visa status, can shape an individual’s ability and willingness to pursue entrepreneurial activities. Further research is needed to examine how these factors interact and affect training experiences and entrepreneurial outcomes.

This work was supported by the National Science Foundation I-Corps Hubs Program (grant number 2048612). The dataset “Impact of NSF’s I-Corps National Program on Individual Participants” was prepared by VentureWell.